Some important thermal parameters of thermal power units are difficult to measure directly

to solve this problem

a soft-measurement method based on correlation vector machine(RVM) was proposed. In this method

the combined kernel function replaces the conventional single kernel function

the improved whale optimization algorithm(IWOA) is used to optimize the kernel function parameters and the combined kernel weight coefficient

and the IWOA-RVM soft measurement model is established by using the optimized parameters and the actual operating conditions data of the unit. Moreover

the sample update strategy is applied to realize online update of the model. Furthermore

the above model was employed to the parameter measurement of a 600 MW coal-fired boiler.The results show that

by using the IWOA-RVM soft measurement model

the root mean square error of the predicted oxygen content in flue gas and carbon content in fly ash is 0.085 2 and 0.088 4

both are lower than that by the BP neural network and support vector machine soft measurement model. The proposed model requires less correlation vector

and has good model sparseness

high test accuracy and strong generalization ability.